Currently, security surveillance of unguarded sites for burglary, vandalism or other unlawful activity, or for augmenting protection of guarded sites, usually involves taking video camera images of site activity and transmitting the images to a remote central monitoring point. Detection of an actionable incident or event at the site is left to a human observer, who periodically inspects an array of video displays of multiple sites. Because of fatigue and other factors, the chances of the observer failing to notice an incident are substantial. Further, automated reading of video imagery and signal patterns to provide discrimination among physical objects and their movement is difficult because of the complex and expensive signal processing required to accurately discriminate discrete object movement activity in a video pattern. Also, video bandwidth limitations of voice-grade circuits restrict the information that can be transmitted as video images.